Does market power in Search lead to higher final product prices?
The CMA suggest that market power in General Search could lead to higher prices of products and services that use Search ads.
This didn’t make sense to me — Search ad prices are set by an auction, and I couldn’t see how market power in General Search would affect the auction price.
After more thought, I think the CMA suggestion is theoretically possible, but it relies on consumers choosing General Search providers based on the quality of their adverts, including prices; and General Search providers reacting to this with deep integration with advertisers to understand prices and consumer behaviour. I think both of these are highly fanciful in practice.
The CMA concern
Figure 2.7 in the CMA Digital Advertising interim report sets out four potential “theories of harm”. The second theory suggested that “consumers could pay higher prices for goods and services across the economy”. This struck me as a pretty serious suggestion, but I just couldn’t understand it.
The theory seems to go: Concentration in Search or Social Display → higher advert prices → higher product prices.
In Search, the price for an ad is set by an auction and only paid if the ad is clicked on. Advertisers pay what the lead is worth to them. I just couldn’t see how the market power of the search engine affected that.
The CMA have a go at explaining their thinking at paragraph 58:
Best foot forward
The CMA therefore suggest five possible ways that market power in General Search might translate to increased advertising prices:
- changing the number of ads,
- changing how the ads are presented,
- changing how relevance is assessed,
- changing reserve prices, and
- changing the way that the matching algorithm works.
Changing the number or presentation of ads might change the price by changing the value of the ads. But so long as the auction mechanism is unchanged then it won’t change the price-per-conversion, which is what matters.
As I understand it, reserve prices are used primarily to reduce the number of low-quality adverts shown, and to remove click-bait or irrelevant adverts. It is unclear how lowering them will benefit consumers.
The most compelling CMA theory, for me, is that competition on the user side of the market might lead General Search providers to compete on the price of the goods advertised on their service. This competition would be realised by changing the relevance and matching algorithms to forego ad revenue on the advertising side of the market in exchange for lower prices to consumers on the consumer side of the market.
But does it stand?
For increased competition between General Search providers to lead to lower prices for advertised products, I think it requires the following four conditions:
- An important element of consumer preferences when choosing between General Search providers is the quality of adverts, including the relative price of goods advertised.
- This consumer preference leads General Search providers to place a higher ranking on consumer relevance when choosing adverts than present, at the expense of other indicators such as the value to the advertiser — noting that the value to the advertiser is itself a reasonable proxy for consumer relevance since it indicates the likelihood of a transaction.
- General Search providers are able to incorporate a reliable measurement of relative price in their assessment of relevance.
- This isn’t already happening.
On (1), while it is likely that consumers care about the price of goods advertised, and that advertising reduces the price of goods, I can see no evidence that consumers choose between General Search providers based on further increases in ad quality. Indeed, the CMA report often assumes that ads are “bads”. Principle competitors to Google, such as DuckDuckGo arguably compete on having worse and less relevant ads due to less use of personal data.
On (2) it is hard to see how search engines could over-weigh consumer relevance, as measured by likelihood to click on an ad without incentivising low-quality click-bait adverts. It would involve deep integration with advertisers to understand how consumers behave post-click, integration which may not be appreciated given the concerns raised elsewhere in the CMA report.
There is an equally large practical problem with (3), to the extent that it isn’t done already where possible on specialised search like products. How would a Search provider know if, say GoCompare was generally cheaper than CompareTheMarket? on a genuinely like-for-like basis?
And on (4), while entrant general search services do not appear to differentiate themselves on ad quality it seems likely that specialised services already do. Google and, say, Amazon, appear to me to already compete on the quality and price of products available to buy through their ecosystems.
And this is ignoring the final link, which is that higher advertising prices would lead to higher product prices. Or that online adverts are generally cheaper and more effective than off-line and that this has led to a decrease in overall advertising spend (though of course the counterfactual might be that spend went down even further).
So I’ve shifted from thinking that the theory was just wrong to thinking that it is hypothetically possible but inconsistent with revealed consumer preference and very difficult to imagine in practice.
I can see rents from any market power on the consumer side of the market being returned directly to consumers in some way. But returning them via the advertising side of the market feels convuluted, less likely to attract customers than direct improvements on the consumer side, and therefore fanciful.
I look forward to the CMA (or others) putting more theoretical and empirical flesh on the bones of what at present seems to me to be a serious but pretty flimsy theory of harm.
(I should disclose that I advise Google, in a rather roundabout way, on this case. But, as I hope is obvious, these thoughts are very much my own attempt to get to grips with an intriguing theory of harm).